Skip to content

AGENTS.md

Latest commit

 

History

History
91 lines (62 loc) · 2.93 KB

AGENTS.md

File metadata and controls

91 lines (62 loc) · 2.93 KB
Raw

ROS API CLI

Use the ros command-line tool as the source of truth for every lookup. This skill requires the CLI to be installed locally before use.

Hard Rules

  1. Install ros-cli before using this skill. If the ros command is unavailable, stop and install it before any lookup.
  2. Run the real ros CLI. Do not fabricate results.
  3. Run the onboarding checks after installation and before the first query in a session, or whenever the CLI or config may have changed.
  4. Prefer --output compact or --output table for user-facing summaries. Use --output json only when the raw payload is needed.
  5. Keep API keys out of chat when possible.
  6. Treat returned DOI, ISBN, and SHA256 values as authoritative identifiers.

Onboarding

  1. Install or upgrade the CLI, then verify it is available:

    pip install --upgrade git+https://github.com/Shannon4Science/ros-cli.git
ros --version

    If installation is blocked in the current environment, tell the user the skill cannot be used until ros is installed.

  2. Check configuration:

    ros config show

    If the API key is missing, direct the user to: https://docs.ros.shlab.tech:18443/concepts/authentication/#get-api-key

    Then configure the CLI with either:

    ros config set-key YOUR_API_KEY
ros config init

    Use ros config set-url URL only when the user needs a non-default base URL.

  3. Verify connectivity:

    ros metadata query --search "title:test" --page-size 1 --output compact

Standard Workflows

Search metadata:

ros metadata query --search "title:machine learning" --output compact
ros metadata query --search "abstract:reinforcement learning" --fields "title,author,doi,publication_published_year" --page-size 10 --sort "publication_published_year:desc" --output table

Fetch one metadata record:

ros metadata fetch --doi "10.1590/1806-9126-rbef-2022-0101" --fields "title,author,abstract,doi"
ros metadata fetch --isbn "9780262046824"

Batch fetch metadata:

ros metadata batch-fetch --ids "[{\"field\":\"doi\",\"value\":\"10.1234/a\"},{\"field\":\"isbn13\",\"value\":\"9780262046824\"}]" --fields "title,author,doi"
ros metadata batch-fetch --ids-file identifiers.json

Search content:

ros content query --search "title:deep learning" --fields "sha256,title,file_format,content_url" --output table

Fetch content by SHA256:

ros content fetch --sha256 "dee1a64db5c1117b044f945abdd371179119e63f22dc8854bbf2b0427649a204" --fields "sha256,title,content_url"

Full-Text Retrieval Pattern

When the user starts from a paper title, DOI, or ISBN and needs the full extracted text:

  1. Query or fetch metadata and request access_xinghe_repository_sha256.
  2. Select the relevant SHA256 hash from that field.
  3. Fetch the content record with ros content fetch --sha256 ....
  4. Use the returned content_url as the time-limited link to the extracted text.